perfsonar development work

[Wenji Wu <>]

> Flapping occurs at both the IGP and BGP levels. MRAI is not relevant to
> a whole class of problems that introduce route flapping, especially with
> protocols that are trying to achieve sub second, or sub 60 ms  
> convergence.

Yes, flapping occurs at both the IGP and BGP levels. On my way to lunch, I 
still think about your comment "There are several known causes for persistent 
route flapping on short timescales". 

Could you give me the details on what time scale? 

My thought is: even on short timescales, the time scale would be much larger 
than the load balancing. Router needs to send/receive router updates, it also 
need to recalculate the routing table, and populate the updated routing table 
to the forwarding tables. Even a specific routing protocol converge really 
fast. In real implementation, routing table's recalculating periodicity is 
not at the same time scale as load balancing. I do not have much experience 
on router management. Not sure how often the routing table would be 
recalculated. But still I think the number won't be small, otherwise, it will 
cause network unstable. For BGP, the default MARI is 30s. So i also imagine 
the routing table's recalculating periodicity would be at second level. So, I 
imagine that the time scale of the routing flaps is still much higher than 
load balancing. 

when using OWAMP, we might send out bursty packets, either to mimic the tcp 
burstiness, or probe the intra-flow load balancing. Intra-flow load balancing 
might cause different TTL for the consecutive packets in the same burst. 

> 
> Not necessarily.  Routes can flap between 2 good paths with no packet  
> loss.

In distributed context, route flaps might cause instantaneous 
disconnectivity, even there are good paths, because the routers are not all 
synchronized immediately at the same time. packets are dropped by 
disconnectivity, not bad paths. But of cause, there are exceptions, depending 
on the causes of route flaps.

wenji